Many of those new to R remain stuck in the peaceful world of base R graphics, woefully unaware of the wonderful world of visualization lying just a few taps of the keyboard away!

The goal of this post is to describe the visualization world that lives within the R universe (and a little beyond). I will start from the basics of inbuilt R plotting, introduce the (hopefully already familliar) ggplot2, then blast off into the interactive stratosphere!

library(ggplot2)
library(knitr)
library(dplyr)
library(DT)
library(plotly)
library(lubridate)

I will use the Texan house prices dataset (loaded with the plotly package) to describe the visualization ecosystem. Feel free to play around with the table below which displays the dataset.

datatable(txhousing)

Base plot

The first thing you probably learnt about visualization in R is how to produce a simple scatterplot. For example, to plot a static scatterplot of date versus sales, you simply tell R to literally “plot” one against the other.

plot(txhousing$date, txhousing$sales)

The result ends up being fairly ugly. You can make it a little prettier, but it’s still fairly inflexible.

plot(x = txhousing$date, 
     y = txhousing$sales, 
     col = "cornflowerblue", 
     pch = 16,
     xlab = "date",
     ylab = "sales")

Fortunately, a guy named Hadley Wickham came along one day and completely designed his own plotting universe for R. Although it comes with its own language, once you have a handle on the syntax, it is extremely powerful.

ggplot2: enhancing static plots

To produce a similar, but prettier plot to that above, we can use the following:

ggplot(txhousing) +
  geom_line(aes(x = date, 
                y = sales, 
                group = city),
            alpha = 0.4) 
## Warning: Removed 430 rows containing missing values (geom_path).

The syntax can be read as “make a ggplot object from the diamonds dataset then add some points to it”. Initially, this might just seem like a more complicated way to to base plotting, but it’s so much more: the ability to add themes and objects to a plot in a desired order is utterly revolutionary!

Since this post isn’t really about ggplot2, I’ll leave you with a gentle encouragement to use ggplot2 if you don’t already do so. Let’s move onto the even more juicy stuff: interactivity.

Plotly: a first foray into interactivity

Wouldn’t it be cool if, instead of having to learn all this extra fancy stuff to do interactivity, you could just tell R to make your existing plot interactive? Sounds amazing right? Well I’m here to tell you that it is literally that easy; just use plotly to transform your ggplot object into an interactive plotly object.

Your interactive plot will have tooltips, zooming, and panning enabled by default.

# define our ggplot2 plot
g <- ggplot(txhousing) +
  geom_line(aes(x = date, y = sales, group = city),
            alpha = 0.4)
# make it interactive!
ggplotly(g, tooltip = c("city"))

You can also use plotly independently of ggplot2 for even further customization. To convert your data directly into an interactive plotly object

txhousing %>% 
  # group by city
  group_by(city) %>%
  # initiate a plotly object with date on x and median on y
  plot_ly(x = ~date, y = ~median) %>%
  # plots one line per city since p knows city is a grouping variable
  add_lines(alpha = 0.2, 
            name = "Texan Cities", 
            hoverinfo = "none") %>%
  add_lines(name = "Houston", 
            data = filter(txhousing, city == "Houston")) %>%
  add_lines(name = "Dallas", 
            data = filter(txhousing, city == "Dallas"))

A more flexible way to do this is using the add_fun capabilities

# add the plot for all Texan cities
allCities <- txhousing %>%
  group_by(city) %>%
  plot_ly(x = ~date, y = ~median) %>%
  add_lines(alpha = 0.2, name = "Texan Cities", hoverinfo = "none")

# define a reusable function for highlighting a particular city
layer_city <- function(plot, name) {
  plot %>% filter(city == name) %>% add_lines(name = name)
}

# add the individual plots
allCities %>% 
  # add a layer for Houston
  add_fun(layer_city, "Houston") %>%
  add_fun(layer_city, "Dallas")

Adding a slider

allCities %>% 
  # add a layer for Houston
  add_fun(layer_city, "Houston") %>%
  # add a layer for Dallas
  add_fun(layer_city, "Dallas") %>%
  rangeslider

Linking with Crosstalk

As far as ggplotly() and plot_ly() are concerned, SharedData object(s) act just like a data frame, but with a special key attribute attached to graphical elements.

library(crosstalk)
library(htmltools)
# define a SharedData object, grouping by year
sd <- SharedData$new(txhousing, ~city)

# plot median house prices by month, with one line per year
p <- ggplot(sd, aes(date, median)) +
  geom_line() 

# turn plot into a plotly object
gg <- ggplotly(p, tooltip = "city")

# highlight options
highlight(gg, on = "plotly_click", dynamic = TRUE)
highlight(gg, on = "plotly_click", dynamic = TRUE, persistent = TRUE)

shiny

If you’re looking for further interactive options, and want to create a solid web-based application, then you need look no further than the shiny package, which allows you to create and host your own interactive application.

d3 in R